Is it not feasible to simply raise the bar on what is expected in a graded essay? Or is the human/chatgpt combo no better than just chatgpt?